In silico drug repurposing and in vitro validation of cestode fatty acid binding proteins

This study utilized an integrated in silico screening strategy to identify and experimentally validate hydrochlorothiazide, naratriptan, fenticonazole, and montelukast as potential repurposed drugs targeting cestode fatty acid binding proteins for the treatment of echinococcosis.

The Gist

The Big Picture: A Digital Treasure Hunt for a Parasite Cure

Imagine a tiny, sneaky thief (a parasite called Echinococcus) that lives inside humans and animals. This thief steals the host's food, specifically fats, to survive. But here's the catch: the thief has no kitchen of its own. It can't cook its own fat; it has to raid the pantry of the person it's living in.

To do this, the thief uses special "backpacks" called Fatty Acid Binding Proteins (FABPs). These backpacks grab the fat and carry it around the thief's body. If you can break these backpacks, the thief starves and dies.

The problem is, the current medicines we have to kill this thief are old, sometimes toxic, and don't work well for everyone. The scientists in this paper wanted to find a new, better weapon. But instead of testing millions of chemicals in a lab (which takes years and costs a fortune), they used a super-fast digital search to find the best candidates, and then tested just a few in the real world.

---

Step 1: The Digital Detective Work (Virtual Screening)

The researchers acted like detectives using two different strategies to find a key that fits the thief's backpack lock.

Strategy A: The "Look-Alike" Search (Ligand-Based)
Imagine you have a photo of a key that you know opens a similar lock. You don't need to see the lock itself; you just need to find other keys that look exactly like your photo.

• The scientists took a list of known "fat-blocking" keys and asked a computer to scan a database of 435,000 different keys (drugs).
• They built a "smart filter" (a mathematical model) that learned the shape and features of the good keys.
• The Result: The computer filtered out almost everything and said, "Hey, these three drugs look like they might work!" (Hydrochlorothiazide, Naratriptan, and Fenticonazole).

Strategy B: The "3D Puzzle" Search (Structure-Based)
This time, instead of looking at keys, the scientists looked at the lock itself (the FABP protein). They built a 3D digital model of the lock.

• They noticed that inside the lock, there were tiny "water bubbles" sitting in specific spots. These bubbles were actually helping the lock stay open.
• They told the computer: "When you try to fit a key in, make sure it pushes against these water bubbles."
• The Result: The computer scanned the same 435,000 drugs and found one new candidate (Montelukast) that fit perfectly with the water bubbles.

The Analogy: Think of it like trying to find a specific person in a crowd of 435,000.

• Strategy A says: "Find anyone who looks like a tall guy with a red hat."
• Strategy B says: "Find anyone who fits perfectly into this specific chair in the room."
• By doing both, they found a shortlist of four suspects to investigate further.

---

Step 2: The "Drug Repurposing" Twist

Here is the clever part. The four drugs they picked weren't brand-new inventions. They were old, approved drugs already used for other things:

• Hydrochlorothiazide: Used for high blood pressure.
• Naratriptan: Used for migraines.
• Fenticonazole: Used for fungal infections.
• Montelukast: Used for asthma.

Why is this cool?
Imagine you have a key that opens a front door. You realize it also opens a back door you didn't know about. You don't need to build a new key factory or wait 10 years for safety tests. You already have the key! This is called Drug Repurposing. It's faster, cheaper, and safer because we already know these drugs are safe for humans.

---

Step 3: The Real-World Test (In Vitro Validation)

The computer said, "These four drugs should work." But computers can make mistakes. So, the scientists went into the lab to prove it.

They grew the parasite's "backpacks" (FABPs) in a petri dish and added the drugs. They used a special glowing dye (ANS) that changes color when the backpack is full of fat.

• The Test: If a drug works, it kicks the fat out of the backpack, and the glow changes.
• The Result:
  • Hydrochlorothiazide (the blood pressure pill) worked! It successfully grabbed onto the parasite's backpacks and blocked them.
  • The other three drugs were harder to test because they interfered with the glowing light, but the computer predictions were strong enough to keep them on the list for future study.

---

The Conclusion: Why This Matters

This paper is a success story of modern science.

1. Efficiency: Instead of testing 435,000 drugs one by one (which would take a lifetime), they used math to narrow it down to 4 in a fraction of the time.
2. Cost: They found a potential cure for a disease that mostly affects poor regions (Neglected Tropical Diseases) without spending millions on new drug development.
3. Hope: They found that a common blood pressure pill might actually kill a dangerous parasite. This gives doctors a new tool to fight a disease that has been hard to treat.

In a nutshell: The scientists used a digital "sieve" to filter a massive ocean of drugs, found a few that looked like they fit the parasite's lock, and proved in the lab that one of them (a blood pressure med) actually works. It's a smart, fast, and cheap way to save lives.

Technical Summary

1. Problem Statement

Echinococcosis, caused by Echinococcus granulosus (cystic) and Echinococcus multilocularis (alveolar), is a neglected tropical disease (NTD) with significant public health burdens. Current treatments rely on benzimidazoles and praziquantel, which suffer from poor tolerability, variable efficacy depending on cyst characteristics, and limited availability.

• Biological Target: Cestodes lack de novo fatty acid and cholesterol biosynthesis pathways, making them dependent on host lipids. Fatty Acid Binding Proteins (FABPs) are critical for lipid trafficking in these parasites and are highly expressed, making them promising drug targets.
• Challenge: There is a lack of new anthelmintic agents. Traditional experimental screening is costly and time-consuming. There is a need for efficient strategies to identify novel inhibitors, particularly through drug repurposing (using existing approved drugs) to accelerate development.

2. Methodology

The study employed an integrated virtual screening (VS) strategy combining ligand-based and structure-based approaches to screen ~435,000 compounds, followed by in vitro validation.

A. Ligand-Based Virtual Screening (LBVS)

• Dataset Construction: A dataset of 288 compounds was compiled based on activity against human FABP4 (AFABP), assuming target repositioning potential.
  • Active: 187 compounds (IC₅₀/Kᵢ ≤ 10 µM).
  • Inactive: 101 compounds (IC₅₀/Kᵢ > 20 µM).
  • Decoys: Generated using LUDe and DUDE-Z tools to create retrospective libraries (RL1: 7,699 compounds; RL2: 5,051 compounds).
• Modeling:
  • 4,000 linear classifiers were generated using Mordred descriptors and a random subspace approach.
  • Ensemble Learning: Individual models were combined using five operators (MIN, PROD, AVE, VOTE, RANK). The MIN operator with a 32-model ensemble (MIN32) was selected based on the best BEDROC (Boltzmann-enhanced discrimination of ROC) score, indicating superior early enrichment.
  • Validation: Models were validated via Y-randomization and Leave-Group-Out (LGO) cross-validation to ensure robustness.
• Threshold Selection: A Positive Predictive Value (PPV) surface analysis was used to select a cutoff (0.400) that balanced sensitivity and specificity, prioritizing high specificity to reduce false positives in the prospective screen.

B. Structure-Based Virtual Screening (SBVS)

• Target: Crystal structure of E. granulosus FABP1 (PDB ID: 1O8V).
• Water Site Identification: 100-ns Molecular Dynamics (MD) simulations were performed on the apo-protein. The WATCLUST algorithm identified three distinct water clusters (WS) near the binding triad (Arg107, Arg127, Tyr129) with high occupancy probabilities.
• Docking Protocol:
  • Four docking methods were evaluated: AutoDock-GPU, AutoDock Vina, Vinardo, and AutoDock Bias (AD-Bias).
  • Biasing: The identified water sites were modeled as hydrogen-bond donor (D) or acceptor (A) biases.
  • Optimization: The AD-Bias (-D-) protocol (utilizing a single donor bias at the central water site, WS2) yielded the highest AUC-ROC values (0.779 for dRL1, 0.905 for dRL2).
• Screening: The same chemical libraries used in LBVS were screened using the optimized AD-Bias protocol.

C. Experimental Validation

• Selected Hits: Four compounds were prioritized for in vitro testing based on repurposing potential, commercial availability, and cost:
  1. Hydrochlorothiazide (LBVS hit)
  2. Fenticonazole (LBVS hit)
  3. Naratriptan (LBVS hit)
  4. Montelukast (SBVS hit)
• Assay: ANS (1-anilinonaphthalene-8-sulfonic acid) fluorescence displacement assays were performed against recombinant EgFABP1, EmFABP1, and EmFABP3. Oleic acid served as the positive control.
• Metrics: Apparent dissociation constants ($K_{d,app}$) were calculated based on fluorescence changes.

3. Key Results

Computational Performance

• LBVS: The MIN32 ensemble achieved high specificity (Sp ≈ 0.994) and a BEDROC of ~0.60. From ~435,000 compounds, 56 hits were identified.
• SBVS: The AD-Bias (-D-) protocol significantly outperformed standard docking. It identified 98 hits, none of which overlapped with the LBVS hits, demonstrating the complementary nature of the two approaches.
• Water Role: The inclusion of water-mediated interactions (specifically the donor bias at WS2) was critical for accurate docking performance, validating the hypothesis that structural water molecules stabilize ligand binding in FABPs.

Experimental Validation

• Hydrochlorothiazide: Successfully validated as a binder to all three tested FABP isoforms.
  • $K_{d,app}$ values: 2.19 µM (EgFABP1), 9.90 µM (EmFABP1), and 2.75 µM (EmFABP3).
  • This confirmed the predictive power of the LBVS workflow.
• Other Compounds: Montelukast, fenticonazole, and naratriptan showed fluorescence interference in the assay, preventing the calculation of precise $K_{d,app}$ values. However, their selection highlights the potential of the screening pipeline.
• Specificity: The study confirmed that EgFABP1 has higher affinity for hydrochlorothiazide compared to the Em isoforms, while EmFABP3 showed the highest affinity for the natural ligand (oleic acid).

4. Key Contributions

1. Integrated Screening Pipeline: Demonstrated the efficacy of combining ligand-based (ensemble modeling) and structure-based (water-aware docking) virtual screening to identify novel binders for NTD targets.
2. Water-Mediated Docking: Validated the use of MD-derived water sites (WATCLUST) as biasing potentials in docking, showing that accounting for specific water molecules significantly improves screening accuracy for FABPs.
3. Drug Repurposing Success: Identified hydrochlorothiazide (a thiazide diuretic) as a potent binder to cestode FABPs. This provides a repurposing candidate with established safety profiles, potentially bypassing early-stage toxicity testing.
4. Complementary Hit Discovery: Showed that LBVS and SBVS yield non-overlapping hit lists, suggesting that using both strategies maximizes the chemical space coverage for difficult targets like FABPs.

5. Significance

• Therapeutic Impact: The identification of hydrochlorothiazide as an FABP binder offers a rapid path to developing new treatments for echinococcosis. Repurposing approved drugs reduces development time and cost, which is crucial for NTDs that lack private sector investment.
• Methodological Advancement: The study provides a robust framework for targeting proteins with complex hydration shells. By explicitly modeling water networks, researchers can overcome limitations of traditional docking, leading to higher hit rates and better predictive performance.
• Future Directions: The confirmed affinity of hydrochlorothiazide serves as a starting point for hit-to-lead optimization to develop specific, high-affinity anthelmintics. The study also highlights the need for alternative assay methods for compounds with spectroscopic interference to fully validate all in silico hits.